摘要 高分辨率遥感影像面向对象信息提取的前提和关键步骤是影像分割。在分析水体信息提取方法的基础上,提出多尺度-光谱差异分割的方法。首先,以河南省开封市某区域为例,采用多尺度分割的方法对GF-2影像进行分割,确定最优分割尺度;然后,在保证地类分割纯度的基础上,对分割的影像进行光谱差异分割,得到研究区域水体分割最佳效果;最后,利用水体在近红外波段均值的特征,设置阈值提取到满意的水体信息结果。根据目视评价和定量评价结果,证明该方法在研究区域是可行的。 The prerequisites and key step of object⁃oriented information extraction high⁃resolution remote sensing image is image segmenta⁃tion.based on the analysis of previous water information extraction methods,a new multiscale⁃spectral differential segmentation method was proposed.Firstly,taking an area in Kaifeng City of Henan Province as an example,the multi⁃scale segmentation method was used to segment the GF-2 image and determine the optimal segmentation scale;then,on the basis of ensuring the purity of land⁃cover category segmentation,the spectral differential segmentation was performed on the segmented images to obtain the best effect of water segmentation in the study area;finally,according to the characteristics of the mean value of water in the near⁃infrared band,a threshold value was set to extract the satisfac⁃tory water information results.based on the visual and quantitative evaluation,the proposed method is proved to be feasible in the study area.
机构地区 黄河水利职业技术学院
出处 《人民黄河》 CAS 北大核心 2021年第5期80-83,90,共5页 Yellow River
基金 国家自然科学基金资助项目(51809110) 河南省科技攻关项目(212102311147)。
关键词 面向对象 GF-2影像 多尺度-光谱差异分割 水体信息提取 object⁃oriented method GF-2 image multiscale⁃spectral differential segmentation water information extraction
